The Development of Sequential Recommendation Systems Using Deep Learning

• Published in: 2023 International Conference on Data Science and Network Security (ICDSNS) pp. 1 - 7
• Main Author: Grover, Madhur
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.07.2023
• Subjects: CNN; Deep learning; Grasping; Network security; Predictive models; RNN and GAN; Self-supervised learning; Sequential Recommendation Systems; Training; Transformers
• DOI: 10.1109/ICDSNS58469.2023.10245109

Recommender systems have shown to be a useful tool for dealing with the overwhelming nature of the information available online. Given its broad implementation in many online applications and its potential ability to mitigate numerous issues connected to over-choice, the value of recommendation systems cannot be emphasized. Not only has deep learning shown impressive performance in recent years, but its ability to learn representations of features from scratch is also appealing to researchers in many other domains. Recent studies on retrieval of data and recommender systems have shown the widespread impact of deep learning. In particular, we provide a new categorization framework for consecutive recommendation tasks, within which we systematically describe exemplary DL-based algorithms for various sequential recommendation situations. In order to analyze user-item interaction with greater precision and adaptability, it is helpful to have a sequential knowledge of user preferences. Because of this, sequential recommendation models improve suggestion quality by easily grasping the relationship between consumers and the things they are interested in, and they also get insight into customer behavior via temporal characteristics. While early versions of RNN and CNN struggled with long-term user-item reliance owing to things like lines, points, and nodes, the transformer-based technique was developed to solve this problem. These issues will be investigated, as will self-supervised learning (SSL)-based solutions that were developed to deal with the data sparsity issues that plague recommender systems.